Google AI Edge Foresight
An experimental meeting companion that Google says runs on fully local AI, taps system audio and the microphone, works with any meeting platform and can run completely offline
Tool Overview
Details, capabilities and usage guidance
Features
- ✓ Fully local: Google says Foresight is powered by fully local AI processing, that sensitive information stays on the device, and that it keeps working without a network connection with no cloud subscription costs
- ✓ Works with any meeting platform: Google says it integrates directly with system audio and the microphone, so it works out of the box anywhere and can run completely offline
- ✓ Enhanced note-taking: Google says it enriches manual shorthand notes with details retrieved in real time from the current meeting conversation
- ✓ Cross-modal retrieval: Google says you can find images, documents, transcripts and notes using natural language
- ✓ Model foundation: Google says it is powered by EmbeddingGemma 2 and Gemma 4, mapping text, audio and visual data into a single vector space for local, secure search over your personal knowledge library
How to Use
- Open Google's official AI Edge Foresight page and download the Mac app, which Google labels an experimental project
- Grant it access to system audio and the microphone as described, so it works with any meeting platform without per-platform setup
- Take shorthand notes during the meeting: Google says Foresight enriches them automatically from the live conversation and can answer questions raised mid-meeting
- Afterwards, search across images, documents, transcripts and notes in natural language, all processed locally on the device
FAQ
What is Google AI Edge Foresight?
On its developer blog Google's AI Edge team describes Foresight as an experimental app for Mac that brings a context-aware meeting companion directly to your device, powered by fully local AI processing, assisting with note-taking, indexing and recalling your conversation transcripts and private files.
Which meeting platforms does it support?
Google says it integrates directly with your system audio and microphone, so it works out of the box with any meeting platform and can even run completely offline - no per-platform integration with Zoom, Meet and similar tools is needed.
How is it different from cloud note-takers?
Google stresses on-device processing: sensitive information stays with you, you keep working without a network connection, and there are no cloud subscription costs. Most meeting note tools, by contrast, send audio to a cloud service.
Which models power it?
Google says Foresight is powered by EmbeddingGemma 2 and Gemma 4 models. The same post describes Google DeepMind's EmbeddingGemma 2, an open-weight multimodal embedding model covering text, vision and audio in a 740M parameter footprint, needing as little as about 191MB active RAM for text-only weights and roughly 567MB for the full multimodal model on a Google Pixel 11 Pro.
What exactly does the note-taking do?
Google lists three capabilities: enriching manual shorthand notes with details retrieved in real time from the current conversation; cross-modal retrieval that finds images, documents, transcripts and notes in natural language; and doing all of it on device, so data stays local, network interruptions do not matter and there are no cloud subscription costs. Google says it maps text, audio and visual data into a single vector space for local, secure search across your personal knowledge library.
How can I try it now?
Google's page offers the Mac app for download and labels the project experimental. The same blog post notes the surrounding ecosystem: Google AI Edge Gallery gained Instant Media Search and Video Moments Finder showcases powered by EmbeddingGemma 2, and the model will also be offered as a service on Android through ML Kit in the coming weeks.